Computational cognitive scientist at Harvard probing how minds and machines reason, learn, and interpret rules
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
21
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
This axis is not yet scored: aggregate publication-output metrics are insufficient in the current data. It would measure the volume, venue mix, and per-year distribution of the researcher's published work, so a full output picture awaits more complete indexing.
Impact—
This axis is not yet scored because citation-based impact aggregates are not compiled here. It would measure scholarly influence through total and per-paper citations, citation velocity, and uptake by follow-up work; no estimate should be inferred from other axes.
Focus0/100
Scores 0 on focus — the most dispersed end of this axis. The field split is exactly even: 0.5 computational cognitive science / AI evaluation and 0.5 NLP / AI interpretability (computational law). Topic mass also spreads across five clusters, led by Child and Animal Learning Development (82), Categorization, Perception and Language (42), Social Robot Interaction and HRI (27), Visual Perception (25), and Logic, Reasoning and Knowledge (25).
Collaboration82/100
Scores 82 on collaboration. The record shows 61 distinct co-authors in the recent window against 68 works over the last three years, plus a stable circle of twelve recurring collaborators — including Lance Ying, Katherine M. Collins, Guangyuan Jiang, and Akshay K. Jagadish — spanning cognitive science, AI evaluation, and AI-interpretability projects, indicating broad, repeat collaboration rather than isolated work.
Activity100/100
Scores 100 on activity. Per-year counts climb from single digits in 2017–2020 (6, 9, 8, 9) to 15 in 2021, 23 in 2023, 22 in 2024, and a peak of 32 in 2025, with 14 already logged for 2026; 68 works in the last three years confirm a steep, sustained acceleration in research pace.
Reproducibility55/100
Scores 55 on reproducibility, based on a reproducibility mean of 55 across the 2 papers Evolve has reviewed so far. Both reviewed papers — the letter-versus-spirit steering study and the CogGym benchmark — are rated medium reproducibility readiness, suggesting methods that are broadly reproducible but still dependent on released code, data, and model access.
Research summary
Tomer D. Ullman is a researcher at Harvard University working at the intersection of computational cognitive science and AI. OpenAlex-indexed bibliometrics show an h-index of 21 and an i10-index of 40, with publishing activity that has accelerated sharply — from single-digit annual counts through the 2010s to 23 works in 2023, 22 in 2024, and 32 in 2025, and 68 works in the most recent three-year window. Recent work splits evenly, 0.5 and 0.5, between computational cognitive science / AI evaluation and NLP / AI interpretability (computational law), with topic mass concentrated in child and animal learning and development. Two 2026 arXiv papers reviewed by Evolve illustrate the range: CogGym, which unifies 258 cognitive experiments into a comparative human–machine benchmark and finds LLM–human alignment scaling with model size yet still far below human reliability; and a study using anchored-LoRA adapters to steer large language models toward the letter or the spirit of the law, with spirit-side effects stronger than letter-side ones. Collaboration metrics are high, with 61 distinct co-authors recently and a recurring circle of twelve collaborators. The output and impact radar axes remain unscored pending citation aggregates.
Collaboration pattern
Collaboration is broad and repeated: 61 distinct co-authors appear in the recent window alongside 68 works over three years, with a stable core of twelve recurring collaborators — including Lance Ying, Jinzhou Wu, Katherine M. Collins, Guangyuan Jiang, Akshay K. Jagadish, and Sean Dae Houlihan. The pattern suggests many small-team projects spanning cognitive-science benchmarking and AI-alignment work rather than a single long-running lab line.
Research keywords
Highlights
CogGym: a 258-experiment benchmark for comparing human and machine cognition
Steering LLMs toward the letter or the spirit of the law
Timeline
- 2009–2016Earliest indexed years show steady low-volume publishing — 2, 2, 2, 5, 2, 3, 1, and 6 works per year — before the later acceleration.
- 2017–2020Annual counts rise from the early baseline to 9, 8, 9, and 9 works per year.
- 2021–2023Publishing accelerates to 15 works in 2021, 13 in 2022, and 23 in 2023.
- 2024–2025The two strongest indexed years: 22 works in 2024 and a peak of 32 in 2025.
- 202614 works already indexed this year, including two Evolve-reviewed arXiv papers — the CogGym benchmark and the letter-versus-spirit steering study.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric figures (h-index, i10-index, per-year counts, co-author and topic totals) come from OpenAlex only; other databases may differ.
- The output and impact radar axes are unscored because citation aggregates are not yet compiled for this profile.
- Evolve verdicts cover only the two reviewed 2026 arXiv papers and do not represent the full publication record.
- Identity is resolved from public bibliographic data under ORCID 0000-0003-1722-2382; 2026 counts are year-to-date and may change.